Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.pome.sh/llms.txt

Use this file to discover all available pages before exploring further.

The pome CLI is your primary interface to Pome. You use it to run scenarios against a twin, inspect scored results, export run traces, and — when using the hosted control plane — manage sessions and authenticate your account. The same CLI works in both self-hosted and hosted mode; it reads the mode field from your .pome.json config file to know which to use.

​
Install

npm install -g pome

​
Verify the installation

Run the following to confirm the CLI is installed and check for updates: 
pome version
The output shows the installed version and, if a newer release is available, the latest version alongside it: 
pome 1.2.0
Latest: 1.2.0

​
Self-host vs hosted

The CLI works in both modes without any changes to your workflow. When you run pome init, it writes a .pome.json file to your project root with a mode field set to either oss or hosted. The CLI reads this file automatically on every command; you don’t need to pass flags.
  • oss mode — the CLI boots a local twin, runs the evaluator on your machine, and writes trace artifacts to runs/ in your project directory. No network calls to api.pome.sh. No account required.
  • hosted mode — the CLI provisions a twin session via api.pome.sh, runs the evaluator locally against the remote twin, and posts the scored result back to the cloud for storage and dashboard display.

​
Log in to the hosted control plane

If you are using hosted mode, authenticate with your Pome account: 
pome login
This opens a browser window for OAuth. After you authorize, the CLI writes credentials to ~/.pome/credentials.json. For CI environments where a browser is not available, pass your API key directly with the --token flag: 
pome login --token <api-key>
In CI pipelines you can skip pome login entirely by setting the POME_API_KEY environment variable. The CLI reads it automatically and uses it for all authenticated requests. Store the key as a repository secret and expose it in your workflow:
env:
  POME_API_KEY: ${{ secrets.POME_API_KEY }}

​
Next steps

CLI commands

Full reference for every pome command — run, inspect, export, twin, and more.

Environment variables

All environment variables the CLI reads, including LLM endpoint configuration and debug flags.

Configuration

The .pome.json config file — mode, defaults, and per-project overrides.

Quickstart

Not installed yet? Follow the quickstart to boot the twin and run your first scenario.